search

fsdonks / m4

Other
0 stars 1 forks source link

Normalized Dwell allows > 1.0 #25

Closed fs-tom closed 5 years ago

fs-tom commented 5 years ago

Related to #24

During exploration and testing, it turns out we readily compute normalized dwells > 1.0 in some cases. This works fine for the intended purpose (sorting by normalized dwell, as a proxy for readiness), but we should look out for possible edge cases. So far, regarding #24, we're trying to incorporate optional policy-defined finite cycle length goals as a companion for both normalized dwell, and for policy transfers.

fs-tom commented 5 years ago

Looks harmless. The only thing we need it for is comparison, which is okay to have numbers outside the range [0 1]. It's NOT okay for that to exist when doing policy changes to compute a normalized transfer coordinate though.